In recent years, the use of copyright protected works to train artificial intelligence models has generated intense legal and ethical debate. However, the limits of this practice remain diffuse. Now, a new movement in the courts has given a respite to the great technological ones, tilting the balance in their favor.
The Victory (partial) of Anthropic. This week, the company behind Chatbot Claude got a favorable failure. As Reuters points out, a Federal Judge of the US dismissed the request of several musical editors, including Universal Music Group, who sought to provisionally block the use of copyright protected letters in the training of AI models.
The origin of the case. In October 2023, Concord, Abkco Music & Records, Universal Music and several of his subsidiaries led Anthropic to the courts. They accused her of using letters protected by copyright to train Claude, her chatbot of AI, who was able to generate answers with textual or almost textual fragments of her works.
The plaintiffs argued that Anthropic had violated the copyright of at least 500 songs, although some specifically cited, such as Katy Perry’s ‘Roar’, ‘I Will Survive’ by Gloria Gaynor and ‘You Can’t Always Get What You Want’ of the Rolling Stones. They also stressed that the company did not have any license to use those letters.
“There are already several aggregators of song lyrics and websites that fulfill this same function, but these sites have duly licensed the works protected by the copyright of the publishers to provide this service,” said the demand. Here presumably referred to web catalogs such as Genius or Musixmatch.
The judge says ‘no’ to the motion. Judge Eumi Lee rejected the precautionary measure requested by the record records. He considered that they did not demonstrate irreparable damage or clearly evidenced how the use of protected letters to train Claude affected his reputation or the license market. In addition, he stressed that the petition was too wide.
Claude no longer spits song lyrics. If you ask Claude today to recite the letter of ‘You Can’t Always Get What You Want’ of the Rolling Stones or ‘Roar’ of Katy Perry, the Anthropic chatbot will refuse for copyright reasons. Instead, it will offer some data related to the song, the album or the artist, but without reproducing it.
This is because Anthropic reached an agreement with record records to apply filters that prevent their AI models, both current and future, to generate answers that violate copyright. That commitment was, in fact, one of the elements that the judge took into account by rejecting the precautionary measure requested by the editors.
Everything is said. The judicial decision known this week does not end the case. The process continues its course, with both parties trying to achieve a victory. While the record records seek to protect the catalog of the artists they represent, Anthropic defends their training model supporting the legitimate use.
Imágenes | freepik | Anthropic
In WorldOfSoftware | Millions of people are interested again in Chatgpt. The problem is that he has achieved it by violating copyright
